Assessing the medical costs of injury using linked anonymised datasets. The impossible is often the untried

Size: px

Start display at page:

Download "Assessing the medical costs of injury using linked anonymised datasets. The impossible is often the untried"

Kelley Atkinson
8 years ago
Views:

1 Assessing the medical costs of injury using linked anonymised datasets The impossible is often the untried

2 Content Aim of study Scope for study Material and methods Plan of action Research questions There is no crisis to which academics will not respond with a seminar

3 Aim of study utilise large scale anonymised datasets, linked via unique identifiers, to estimate the medical costs of injury incurred by various sectors within the health care system! Goals are dreams with deadlines!

medical costs of injury incurred by various sectors

4 Scope for study So what! There are plenty of other cost of injury studies out there! Big deal!

5 Scope for study Reason 1: Current problems confidentiality constraints prevent researchers from accessing data focus on single/small number of health care sectors patients journey through entire health sector can t be traced over time

on single/small number of health care sectors patients

6 Scope for study Reason 2: Ready made solution advances in anonymous record linkage techniques Multiple datasets can be analysed without contravening confidentiality longitudinal studies incorporating a plethora of health service sectors can be performed

analysed without contravening confidentiality longitudinal

7 Scope for study Reason 3: Changes in health care utilisation continued advances in medical care more survivors of major trauma increasingly long programme of recovery

8 Materials and methods Design longitudinal; retrospective; quantitative Setting Swansea Data sources - Administrative Register (AR) study population - All Wales Injury Surveillance System (AWISS) A&E treatment - Patient Episode Database for Wales (PEDW) hospital care - Outpatient rehabilitative care - GP primary care

Wales Injury Surveillance System (AWISS) A&E treatment - Patient Episode

9 Plan of action 1. Search - A&E / hospital; 01/04/ /03/ Identify - electronic cohort of injured/non-injured individuals 3. Follow-up - A&E / hospital / outpatients / GP; treatment & recovery phases; 1 year pre-injury / post-injury up to 31/03/ Calculate - direct medical costs; stratify by age / gender / mechanism / type / social class 5. Compare/contrast - injured/non-injured cohorts In theory there is no difference between theory and practice. In practice there is

Follow-up - A&E / hospital / outpatients / GP; treatment & recovery phases; 1 year pre-injury / post-injury up to

10 Plan of action Consideration 1 alternative stages of injury Pre - injury period Index - injury period Post injury period The study population will be grouped and analysed separately depending on whether they have sustained an injury in the pre-/postinjury periods

2007 The study population will be grouped and analysed separately

11 Plan of action Consideration 1- alternative stages of injury (continued) - Injury events (A&E; hospital) occurring within X days of each other assumed to be related to the same injury - Research question 2 - Alternative post-injury cut-offs depending on presence of additional A&E/hospital events during follow-up period

be related to the same injury - Research question 2 - Alternative post-injury

12 Plan of action Consideration 2 potential confounders - Injury severity / type - Pre injury health care utilisation - Co-morbidities - Demographics Solution: match injury group to cohort of non-injured individuals in terms of the confounding variable

Co-morbidities - Demographics Solution: match injury group to

13 Plan of action Consideration 3 calculating costs - direct medical costs of injury only - incidence based approach - bottom-up approach - costs defined as the product of incidence, transition probabilities, health care use and unit costs (EUROCOST) unit costs from Trust Financial Return 2 (TFR2) accounts

as the product of incidence, transition probabilities, health care use and

14 Research questions Do the direct medical costs of injury reported following record linkage differ from those based on the separate parallel analysis of unconnected health datasets? How many days should be allowed between 2 hospital admissions before the second can be classed as a new injury event?

differ from those based on the separate parallel analysis of unconnected

15 Research questions 3. Are injury related health care contacts occurring multiple years apart connected with each other? 4. Does accounting for the long-term affects of injury impact upon the scale of the direct medical costs reported?

16 Thank you for listening

Long-term consequence of injury on self-rated health

Long-term consequence of injury on self-rated health Bjarne Laursen SAFETY2010, London September 23, 2010 Bjarne Laursen, Hanne Møller, Anne Mette Hornbek Toft National Institute of Public Health Background